[arin-ppml] Policy Proposal: Open Access To IPv6

Tom Vest tvest at pch.net
Sun May 31 17:40:20 EDT 2009

On May 31, 2009, at 3:29 PM, Paul Vixie wrote:

> tvest at pch.net (Tom Vest) writes:
>> If I remember correctly, creating a swamp only serves to constantly
>> remind those who are stuck with it afterward that swamp creation  
>> was / is
>> a very bad idea. Besides, if you have an idea of where/how one might
>> build a more "solid foundation", persuading us now, up front might  
>> be a
>> more effective way of bringing people around than intentionally  
>> degrading
>> the only "ground" that's currently apparent.
> there are more than two visions (pure hierarchy and pure swamp).   
> for example:
> neighborhood or metro-area mesh networking where local cheap  
> highspeed ISO-L2
> is used to glue geographies together in a way that no telco or  
> backbone net
> is involved... would make better use of available glass and silicon  
> than the
> pure hierarchical model IETF CIDR gave us.  this sounds like a bad  
> idea since
> it would either mean a global swamp (everybody's /56 in the core) or  
> monopoly
> status for incumbants (everybody's /56 came from the same /32) or  
> mass route
> pollution (everybody's /56 becomes a metro-area cutout).
> but what if multihoming was automatic and universal and robust?   
> could a metro
> or neighborhood get unrouteable / non-global IPv6 space for an ISO- 
> L2 overlay
> made up of a hairball of private wireless, private wire, private  
> fiber, and
> automatically use those addresses when talking to reachable  
> endpoints?  (this
> would require something better than RIPv2, so don't try it at home  
> today!)
> or what if a metropolitan connectivity authority wanted to get an  
> IPv6 block
> for all of its FTTH and mobile/wireless endpoints, and rather than  
> buying
> transit for this block, they set up a market of cooperating backbone  
> operators
> and consumers, doing IP-in-IP to deliver global reachability?  (this  
> is like
> what some 802 networks do today but wide area bridging does not  
> scale well.)
> i'm not proposing either of these, not exactly.  i'm saying there  
> ought to
> be room in the RIR allocation policy framework for addressing models  
> that
> are not dreamt of by those who love swamps and those who fear swamps.
> -- 
> Paul Vixie
> __________

Hi Paul,

I think all of these ideas sound great -- esp. the automatic/universal/ 
robust multihoming. Certainly the two scenarios sound like they might  
provide  solid "justification" for the allocation of unique IP address  
resources, given the previous condition (and assuming that the  
aspiring metro/neighborhood operator is capable of providing some  
other evidence of their capability to use the IPv6 addresses in  
pursuit of the stated goals). I am all for creating the conditions  
necessary to enable future innovation -- including esp. the condition  
of open, non-discriminatory, non-adversarial access to "useful" IP  
addresses for those who can put them to (potentially) good use.  
However, I also feel strongly that the duty of "stewardship" requires  
that any relaxation of current allocation policies to accommodate  
hypothetical future innovations should be bounded, at least, by some  
consideration of the most obvious hypothetical future risks.

I'm not 100% on IETF history, but I have a sense that the abandonment  
of backwards compatibility as a requirement for what became IPv6 was  
justified in part by similar arguments, i.e., that encumbering NGN  
addressing with the requirement of consistency/interoperability with  
the known limitations of addressing and routing c. 1992 would entail  
giving up too many lines of anticipated/hypothetical future  
innovation. In hindsight, I think it's probably safe to say that that  
was a net unfortunate choice.

FWIW, I thought that Bill's "mostly open" counter-proposal was  
interesting, enough at least to send him some clarifying questions off- 
list. Bottom line: I think that allocation policies should add zero  
additional barrier to any credible, aspiring IPv6-based service  
implementer. Future scalability risks should be given all due  
consideration, but they should not impede credible efforts to deploy  
now. However, in my mind "needs based" allocation (eligibility +  
quantity) rules fully satisfy this test: for anyone who actually  
possesses the other requisite inputs and plan to deploy service,  
merely demonstrating that fact represents no real additional impediment.


More information about the ARIN-PPML mailing list